Ever since ChatGPT (and the underlying LLM) appeared, many of us have wondered when an LLM-powered chat interface could be integrated with applications we use every day. Take a CRM application, for instance. Wouldn't it be great to use the same chat interface to log leads, update contacts, or make call notes?
With the advent of Model Context Protocol (MCP—yes, another TLA)—think of it as a universal connector for LLM-to-application integrations—this has become not only possible but also fairly straightforward to implement (with a few caveats).
As an experiment, I connected Claude Desktop to a CRM test application via an MCP Server. Whether performing natural language searches, summarizing contact relationships, logging calls, or even cleaning data, the results are seriously impressive.
You still need to know what you're doing, but the chat interface we've become so comfortable with feels natural—like asking someone to log a call or run a report for you.
It's only a matter of time before the venerable but clunky web form gives way to this intuitive interface.
(Em dashes are my own.)
Zaheer Ismail